AI-Driven Tuberculosis Prediction Using Convolutional Autoencoder, Dual-Key Transformer Networks, and Serverless Cloud Computing

Abstract

Tuberculosis (TB) remains one of the leading causes of death worldwide, especially in low-resource settings where access to early diagnostic tools is limited. Recent advancements in artificial intelligence (AI), particularly deep learning techniques such as Convolutional Neural Networks (CNNs), Convolutional Autoencoders (CAE), and transformer-based architectures, have significantly improved automated TB detection from medical imaging data. This paper presents a comprehensive review of emerging trends and challenges in integrating CAE with dual-key transformer networks for smart e-health applications. Additionally, the role of serverless cloud computing in enabling scalable, cost-efficient, and real-time TB prediction systems is examined. Recent studies indicate that hybrid deep learning models combining convolutional and transformer architectures outperform traditional single-model approaches by capturing both local and global features. For example, CNN-transformer hybrid systems have achieved accuracy exceeding 99% in TB detection tasks. Furthermore, transfer learning approaches have demonstrated high classification performance even with limited datasets. Despite these advancements, challenges such as computational complexity, data privacy concerns, and lack of interpretability persist. This review highlights the potential of integrating CAE, dual-key transformers, and serverless cloud computing to develop efficient and intelligent TB prediction systems while identifying key research gaps and future directions.